diff --git a/CHANGELOG.md b/CHANGELOG.md
index da755bc..ed97c54 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -2,6 +2,8 @@
 
 ## [Unreleased]
 
+- Aligns the default local Ollama answer model with the city-core installer
+  pull set by using `gemma4:e4b` when `CIVICCODE_OLLAMA_MODEL` is unset.
 - Hardened the existing CivicClerk ordinance-event intake route with an
   optional shared-intake authorization header for city-core service-to-service
   handoff calls, without broadening trusted staff-header access for the rest of
diff --git a/civiccode/ai_answer.py b/civiccode/ai_answer.py
index 58d1f2e..4ac86d8 100644
--- a/civiccode/ai_answer.py
+++ b/civiccode/ai_answer.py
@@ -31,7 +31,7 @@ def load_local_llm_config() -> LocalLLMConfig | None:
     return LocalLLMConfig(
         provider="ollama",
         base_url=(base_url or "http://127.0.0.1:11434").rstrip("/"),
-        model=os.environ.get("CIVICCODE_OLLAMA_MODEL", "llama3.1:8b").strip() or "llama3.1:8b",
+        model=os.environ.get("CIVICCODE_OLLAMA_MODEL", "gemma4:e4b").strip() or "gemma4:e4b",
         timeout_seconds=float(os.environ.get("CIVICCODE_OLLAMA_TIMEOUT_SECONDS", "20")),
     )
 
diff --git a/tests/test_milestone_7_citation_grounded_qa.py b/tests/test_milestone_7_citation_grounded_qa.py
index 242ef48..93c2857 100644
--- a/tests/test_milestone_7_citation_grounded_qa.py
+++ b/tests/test_milestone_7_citation_grounded_qa.py
@@ -12,6 +12,8 @@
 from conftest import build_suite_staff_headers
 from httpx import ASGITransport, AsyncClient
 
+from civiccode.ai_answer import load_local_llm_config
+
 
 ROOT = Path(__file__).resolve().parents[1]
 
@@ -193,6 +195,17 @@ def log_message(self, format: str, *args: object) -> None:  # noqa: A002
     assert "Source:" in payload["answer"]
 
 
+def test_local_ollama_default_model_matches_city_core_pull(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setenv("CIVICCODE_AI_MODE", "ollama")
+    monkeypatch.delenv("CIVICCODE_OLLAMA_MODEL", raising=False)
+    monkeypatch.delenv("CIVICCODE_OLLAMA_URL", raising=False)
+
+    config = load_local_llm_config()
+
+    assert config is not None
+    assert config.model == "gemma4:e4b"
+
+
 def _ollama_generate_available() -> bool:
     body = json.dumps(
         {